# How hard the land is stocked **Answer.** Uruguay carries **0.856 cattle per hectare** of livestock land (12,265,269 cattle on 14.32 million ha). The densest department, **San José**, carries 1.67; the sparsest of the big grazing departments, Artigas, 0.65. Notebook: https://cow-tools.sophia-labs.com/notebooks/stocking-intensity/stocking-intensity.ipynb · Page: https://cow-tools.sophia-labs.com/notebooks/stocking-intensity/index.html · Data: Uruguay MGAP/SNIG DICOSE 2024–2025 (graph `uruguay-cows` at https://cow-tools.sophia-labs.com/mcp). ## Findings - Uruguay carries **0.856 cattle per hectare** of livestock land (12,265,269 cattle on 14.32 million ha). The densest department, **San José**, carries 1.67; the sparsest of the big grazing departments, Artigas, 0.65. - Density follows the sown pasture: across departments, Spearman ρ = **0.847** between cattle per hectare and the share of land under *praderas*. Nationally 72.4% of declared land is native grassland and only 6.9% sown pasture: intensity is a thin southern veneer. - Of 629 census areas with at least 100 livestock hectares, the densest is area **317003 in Canelones** at 8.33 cattle per hectare; the median area carries 0.91. ## The notebook Most of Uruguay's cattle graze *campo natural*, native grassland that has fed cattle since the 17th century. Where farmers have sown pastures (*praderas*) or improved the native grass, the land carries more animals. This notebook measures stocking density, cattle per hectare of livestock land, at two scales: the 19 departments and the 600-odd *áreas de enumeración*, DIEA's census areas, the finest grain DICOSE publishes. Two inputs, both from the declarations: cattle (beef + dairy, PD_AD) from `datosanimales.csv`, and *superficie ganadera*, the hectares a holder uses for livestock, from `datosgenerales.csv`. ```python # Setup. import altair as alt import pandas as pd import cowtools_client as cw import cowtools_charts as cc names = cc.department_names() ``` ```python # The query: cattle and livestock hectares per census area, 2025. cattle = cw.sparql(""" SELECT ?department ?area (SUM(?n) AS ?cattle) WHERE { GRAPH ?g { ?obs a cow:AnimalFlowObservation ; cow:year "2025"^^xsd:gYear ; cow:department ?department ; cow:enumerationArea ?area ; cow:species ?s ; cow:animalsCount ?n . VALUES ?s { } } } GROUP BY ?department ?area """) land = cw.sparql(""" SELECT ?department ?area (SUM(?ha) AS ?livestock_ha) (SUM(?lu) AS ?livestock_units) WHERE { GRAPH ?g { ?obs a cow:GeneralHoldingObservation ; cow:year "2025"^^xsd:gYear ; cow:department ?department ; cow:enumerationArea ?area ; cow:livestockHectares ?ha ; cow:livestockUnits ?lu . } } GROUP BY ?department ?area """) ae = cattle.merge(land, on=["department", "area"], how="outer") ae["code"] = ae["department"].map(cw.dept_code) ae["area_code"] = ae["area"].map(cw.dept_code) ae["cattle"] = ae["cattle"].fillna(0).astype(int) for c in ("livestock_ha", "livestock_units"): ae[c] = ae[c].apply(lambda v: float(v) if v is not None and v == v else 0.0) print(f"{ae['area_code'].nunique()} census-area codes, {len(ae)} (department, area) cells") ``` ```text 635 census-area codes, 667 (department, area) cells ``` A census area can straddle a department line (the area follows the land; the department follows where the holder is registered), so we key on the pair. Area code `0` collects holdings DICOSE could not place in any area. ```python dept = ae.groupby("code")[["cattle", "livestock_ha", "livestock_units"]].sum().reset_index() dept["name"] = dept["code"].map(names) dept["density"] = dept["cattle"] / dept["livestock_ha"] dept = dept.sort_values("density", ascending=False) national = dept["cattle"].sum() / dept["livestock_ha"].sum() cc.show_table(dept[["name", "density", "cattle", "livestock_ha"]], floatfmt=",.3f", name="density-by-department") ``` | name | density | cattle | livestock_ha | |:---------------|----------:|:----------|:---------------| | San José | 1.67 | 541,254 | 324,103.300 | | Colonia | 1.553 | 537,066 | 345,775.100 | | Canelones | 1.391 | 338,601 | 243,465.000 | | Florida | 1.186 | 1,117,097 | 942,194.800 | | Soriano | 1.15 | 567,452 | 493,507.300 | | Flores | 1.005 | 409,429 | 407,511.500 | | Durazno | 0.97 | 952,171 | 981,607.200 | | Montevideo | 0.918 | 2,736 | 2,980.200 | | Río Negro | 0.875 | 549,369 | 627,763.500 | | Rivera | 0.83 | 639,917 | 770,887.200 | | Cerro Largo | 0.782 | 950,412 | 1,214,639.600 | | Lavalleja | 0.766 | 654,451 | 853,971.200 | | Rocha | 0.755 | 682,644 | 903,944.000 | | Tacuarembó | 0.751 | 1,061,136 | 1,413,868.000 | | Treinta y Tres | 0.725 | 584,734 | 806,167.600 | | Salto | 0.679 | 898,899 | 1,323,035.000 | | Paysandú | 0.679 | 801,512 | 1,180,985.200 | | Maldonado | 0.667 | 234,191 | 351,185.500 | | Artigas | 0.655 | 742,198 | 1,133,714.600 | ### Cross-check against the source ```python g = cw.source_csv(2025, "datosgenerales.csv") a = cw.source_csv(2025, "datosanimales.csv") src_ha = cw.num(g["SuperficieGanadera"]).groupby(g["DepartamentoCodigo"]).sum() bov = a[a["EspecieCodigo"].isin(["1", "11"])] src_cattle = cw.num(bov["AnimalesPorEspeciePD_AD"]).groupby(bov["DepartamentoCodigo"]).sum() d = dept.set_index("code") cw.check("livestock hectares per department: cube = CSV (largest difference, ha)", round(float((d["livestock_ha"] - src_ha.reindex(d.index)).abs().max()), 6), 0.0, source="datosgenerales.csv 2025, SuperficieGanadera") cw.check("cattle per department: cube = CSV (largest difference, heads)", int((d["cattle"] - src_cattle.reindex(d.index)).abs().max()), 0, source="datosanimales.csv 2025") cw.check("national livestock hectares, millions", round(dept["livestock_ha"].sum() / 1e6, 2), 14.32, tol=0.005, source="expedition 2026-07-05: 14.32M ha") cw.check("national density, cattle per livestock ha", round(national, 3), 0.856, tol=0.0005, source="expedition 2026-07-05: 0.856") cw.check("San José density", round(d.loc["16", "density"], 2), 1.67, tol=0.005, source="expedition 2026-07-05") # The cube also materialises this ratio at ingest (cow:metric_bovine_per_livestock_hectare, computed by cow-tools # from its own sums): all 19 departments must agree with ours. derived = cw.sparql(""" SELECT ?department ?value WHERE { GRAPH ?g { ?o a cow:DerivedMetricObservation ; cow:year "2025"^^xsd:gYear ; cow:metric cow:metric_bovine_per_livestock_hectare ; cow:department ?department ; cow:metricValue ?value } } """) derived["code"] = derived["department"].map(cw.dept_code) gap = (d["density"] - derived.set_index("code")["value"].astype(float).reindex(d.index)).abs().max() cw.check("all 19 densities = the cube's derived metric (largest gap)", round(float(gap), 6), 0.0, tol=1e-6, source="cow:metric_bovine_per_livestock_hectare, materialised at ingest") cw.check("Artigas density", round(d.loc["1", "density"], 4), 0.6547, tol=0.00005, source="the July expedition printed 0.66; the exact value is 0.6547, so that was a rounding slip") ``` ```text PASS livestock hectares per department: cube = CSV (largest difference, ha): 0 vs 0 (exactly) — datosgenerales.csv 2025, SuperficieGanadera PASS cattle per department: cube = CSV (largest difference, heads): 0 vs 0 (exactly) — datosanimales.csv 2025 PASS national livestock hectares, millions: 14.32 vs 14.32 (within ±0.005) — expedition 2026-07-05: 14.32M ha PASS national density, cattle per livestock ha: 0.856 vs 0.856 (within ±0.0005) — expedition 2026-07-05: 0.856 PASS San José density: 1.67 vs 1.67 (within ±0.005) — expedition 2026-07-05 ``` ```text PASS all 19 densities = the cube's derived metric (largest gap): 0 vs 0 (within ±1e-06) — cow:metric_bovine_per_livestock_hectare, materialised at ingest PASS Artigas density: 0.6547 vs 0.6547 (within ±5e-05) — the July expedition printed 0.66; the exact value is 0.6547, so that was a rounding slip ``` ```python cw.finding("headline", f"Uruguay carries **{national:.3f} cattle per hectare** of livestock land " f"({int(dept['cattle'].sum()):,} cattle on {dept['livestock_ha'].sum()/1e6:.2f} million ha). " f"The densest department, **{dept.iloc[0]['name']}**, carries {dept.iloc[0]['density']:.2f}; " f"the sparsest of the big grazing departments, {dept.query('livestock_ha > 500000').iloc[-1]['name']}, " f"{dept.query('livestock_ha > 500000').iloc[-1]['density']:.2f}.") ``` > **Finding.** Uruguay carries **0.856 cattle per hectare** of livestock land (12,265,269 cattle on 14.32 million ha). The densest department, **San José**, carries 1.67; the sparsest of the big grazing departments, Artigas, 0.65. ```python cc.show(cc.choropleth(dept, "density", title="Cattle per hectare of livestock land, 2025", subtitle="beef + dairy (PD_AD) ÷ superficie ganadera", legend="cattle per ha", fmt=".2f", domain=[0, 1.8]), "density-map", thumbnail=True, alt_text="Map: stocking density is highest in the south-west (San José, Colonia, Canelones) and lowest in the north-west (Artigas, Salto).") print(cc.map_credit()) ``` ![Map: stocking density is highest in the south-west (San José, Colonia, Canelones) and lowest in the north-west (Artigas, Salto).](charts/density-map.svg) Chart data and encoding: [charts/density-map.vl.json](charts/density-map.vl.json) (Vega-Lite). ```text Map: Natural Earth 1:10m admin-1 (public domain, v5.1.2), joined on the DICOSE department code. Made with Natural Earth. ``` ## Why: the sown pastures The land-use table says how each holding's land is used. Code `1`, *praderas artificiales permanentes*, is the sown, permanent pasture; code `8` is native grassland and stubble. The July 2026 expedition found the department ranking by density follows the share of sown pasture closely (Spearman ρ = 0.847). We recompute it. ```python # The query: hectares by land use per department, 2025. use = cw.sparql(""" SELECT ?department ?use (SUM(?ha) AS ?hectares) WHERE { GRAPH ?g { ?obs a cow:LandUseObservation ; cow:year "2025"^^xsd:gYear ; cow:department ?department ; cow:landUse ?use ; cow:hectares ?ha . } } GROUP BY ?department ?use """) use["code"] = use["department"].map(cw.dept_code) use["use_code"] = use["use"].map(cw.dept_code) use["hectares"] = use["hectares"].astype(float) mix = use.pivot_table(index="code", columns="use_code", values="hectares", aggfunc="sum").fillna(0) shares = mix.div(mix.sum(axis=1), axis=0) dept = dept.merge(shares[["1", "8"]].rename(columns={"1": "praderas", "8": "campo_natural"}), left_on="code", right_index=True) nat = mix.sum() / mix.sum().sum() rho = dept[["density", "praderas"]].rank().corr().iloc[0, 1] cw.check("national share of campo natural", round(nat["8"], 3), 0.724, tol=0.0005, source="expedition 2026-07-05: 72.4%") cw.check("national share of praderas", round(nat["1"], 3), 0.069, tol=0.0005, source="expedition 2026-07-05: 6.9%") cw.check("Spearman ρ, density vs praderas share", round(rho, 3), 0.847, tol=0.0005, source="expedition 2026-07-05") cw.finding("pastures", f"Density follows the sown pasture: across departments, Spearman ρ = **{rho:.3f}** between cattle per " f"hectare and the share of land under *praderas*. Nationally {nat['8']:.1%} of declared land is native " f"grassland and only {nat['1']:.1%} sown pasture: intensity is a thin southern veneer.") ``` ```text PASS national share of campo natural: 0.724 vs 0.724 (within ±0.0005) — expedition 2026-07-05: 72.4% PASS national share of praderas: 0.069 vs 0.069 (within ±0.0005) — expedition 2026-07-05: 6.9% PASS Spearman ρ, density vs praderas share: 0.847 vs 0.847 (within ±0.0005) — expedition 2026-07-05 ``` > **Finding.** Density follows the sown pasture: across departments, Spearman ρ = **0.847** between cattle per hectare and the share of land under *praderas*. Nationally 72.4% of declared land is native grassland and only 6.9% sown pasture: intensity is a thin southern veneer. ```python pts = alt.Chart(dept, title=alt.Title("Density follows the sown pasture", subtitle="one point per department, 2025")).mark_point(filled=True, size=70, color=cc.CATTLE).encode( x=alt.X("praderas:Q", title="share of land under sown pasture (praderas)", axis=alt.Axis(format=".0%")), y=alt.Y("density:Q", title="cattle per livestock hectare"), tooltip=["name:N", alt.Tooltip("praderas:Q", format=".1%"), alt.Tooltip("density:Q", format=".2f")], ) # Label the spread-out departments; the crowded corner is in the tooltip and the table above. labels = alt.Chart(dept).transform_filter( "datum.praderas > 0.09 || indexof(['Artigas', 'Durazno', 'Rivera'], datum.name) >= 0" ).mark_text(align="left", dx=6, dy=-4, fontSize=11).encode(x="praderas:Q", y="density:Q", text="name:N") cc.show((pts + labels).properties(width=520, height=320), "density-vs-praderas", alt_text="Scatter: departments with more sown pasture carry more cattle per hectare; San José and Colonia top right, Artigas and Salto bottom left.") ``` ![Scatter: departments with more sown pasture carry more cattle per hectare; San José and Colonia top right, Artigas and Salto bottom left.](charts/density-vs-praderas.svg) Chart data and encoding: [charts/density-vs-praderas.vl.json](charts/density-vs-praderas.vl.json) (Vega-Lite). ## The census areas Down at the census areas the range is wide. Each dot is one area in its department (areas with under 100 livestock hectares are left out: a few feedlots and dairy yards on tiny land bases would otherwise dominate the scale; they are listed below). ```python # An area can straddle a department line: sum it across departments, and file it under the department # holding most of its livestock land. cells = ae[ae["area_code"] != "0"] home = cells.sort_values("livestock_ha").groupby("area_code").tail(1).set_index("area_code")["code"] areas = cells.groupby("area_code")[["cattle", "livestock_ha"]].sum() areas = areas[areas["livestock_ha"] >= 100].reset_index() areas["name"] = areas["area_code"].map(home).map(names) areas["density"] = areas["cattle"] / areas["livestock_ha"] small = cells.groupby("area_code")[["cattle", "livestock_ha"]].sum().query("livestock_ha < 100 and cattle > 0") order = dept.sort_values("density", ascending=False)["name"].tolist() strip = alt.Chart(areas, title=alt.Title("Cattle per livestock hectare, by census area, 2025", subtitle=f"{len(areas)} areas with at least 100 livestock ha; departments ordered by overall density")).mark_tick( thickness=2, size=14, color=cc.CATTLE, opacity=0.7).encode( x=alt.X("density:Q", title="cattle per livestock hectare", scale=alt.Scale(domain=[0, 4], clamp=True)), y=alt.Y("name:N", sort=order, title=None), tooltip=["name:N", alt.Tooltip("area_code:N", title="census area"), alt.Tooltip("density:Q", format=".2f"), alt.Tooltip("cattle:Q", format=","), alt.Tooltip("livestock_ha:Q", format=",.0f")], ).properties(width=520, height=420) cc.show(strip, "areas-strip", alt_text="Strip plot: each department's census areas as ticks along a density axis; the south-west spreads furthest right.") top = areas.sort_values("density", ascending=False).head(5) cc.show_table(top[["name", "area_code", "density", "cattle", "livestock_ha"]], floatfmt=",.2f", name="densest-areas") print(f"{cells['area_code'].nunique()} census-area codes carry holdings (code 0, unplaced, excluded); " f"{(cells.groupby('area_code')['code'].nunique() > 1).sum()} of them straddle a department line.") print(f"{len(small)} areas with cattle on under 100 livestock ha were left out of the strip plot.") ``` ![Strip plot: each department's census areas as ticks along a density axis; the south-west spreads furthest right.](charts/areas-strip.svg) Chart data and encoding: [charts/areas-strip.vl.json](charts/areas-strip.vl.json) (Vega-Lite). | name | area_code | density | cattle | livestock_ha | |:----------|------------:|----------:|:---------|:---------------| | Canelones | 317003 | 8.33 | 11,589 | 1,391.20 | | San José | 1607005 | 6.38 | 4,879 | 765.10 | | Canelones | 305004 | 5.59 | 4,312 | 771.90 | | Soriano | 1704001 | 4.76 | 18,400 | 3,862.00 | | Florida | 805004 | 4.62 | 20,074 | 4,344.10 | ```text 634 census-area codes carry holdings (code 0, unplaced, excluded); 14 of them straddle a department line. 5 areas with cattle on under 100 livestock ha were left out of the strip plot. ``` ```python cw.check("every area cell is inside a department", int(ae["code"].isna().sum()), 0, source="cube") cw.finding("areas", f"Of {len(areas)} census areas with at least 100 livestock hectares, the densest is area " f"**{top.iloc[0]['area_code']} in {top.iloc[0]['name']}** at {top.iloc[0]['density']:.2f} cattle per hectare; " f"the median area carries {areas['density'].median():.2f}.") ``` ```text PASS every area cell is inside a department: 0 vs 0 (exactly) — cube ``` > **Finding.** Of 629 census areas with at least 100 livestock hectares, the densest is area **317003 in Canelones** at 8.33 cattle per hectare; the median area carries 0.91. ## Where every number came from ```python cc.show_table(cw.checks()[["check", "got", "expected", "rule", "source"]], name="checks") cw.provenance() ``` | check | got | expected | rule | source | |:-----------------------------------------------------------------------|------:|-----------:|:---------------|:-----------------------------------------------------------------------------------------| | livestock hectares per department: cube = CSV (largest difference, ha) | 0 | 0 | exactly | datosgenerales.csv 2025, SuperficieGanadera | | cattle per department: cube = CSV (largest difference, heads) | 0 | 0 | exactly | datosanimales.csv 2025 | | national livestock hectares, millions | 14.32 | 14.32 | within ±0.005 | expedition 2026-07-05: 14.32M ha | | national density, cattle per livestock ha | 0.86 | 0.86 | within ±0.0005 | expedition 2026-07-05: 0.856 | | San José density | 1.67 | 1.67 | within ±0.005 | expedition 2026-07-05 | | all 19 densities = the cube's derived metric (largest gap) | 0 | 0 | within ±1e-06 | cow:metric_bovine_per_livestock_hectare, materialised at ingest | | Artigas density | 0.65 | 0.65 | within ±5e-05 | the July expedition printed 0.66; the exact value is 0.6547, so that was a rounding slip | | national share of campo natural | 0.72 | 0.72 | within ±0.0005 | expedition 2026-07-05: 72.4% | | national share of praderas | 0.07 | 0.07 | within ±0.0005 | expedition 2026-07-05: 6.9% | | Spearman ρ, density vs praderas share | 0.85 | 0.85 | within ±0.0005 | expedition 2026-07-05 | | every area cell is inside a department | 0 | 0 | exactly | cube | ```text kind target rows \ 0 sparql uruguay-cows 660 1 sparql uruguay-cows 667 2 sparql uruguay-cows 19 3 sparql uruguay-cows 152 first line 0 SELECT ?department ?area (SUM(?n) AS ?cattle) 1 SELECT ?department ?area (SUM(?ha) AS ?livesto... 2 SELECT ?department ?value 3 SELECT ?department ?use (SUM(?ha) AS ?hectares) ``` ```python cw.save() ```